I resolved my own issue with this one by changing the dims on beta, alpha, and mu from non_zero_groups to groups and now the groups with 0 successes have mu informed by the population’s data, as I intended. Still sampling well with 0 divergences. I believe I have a basis for moving forward (still open to feedback/suggestions with the graphviz) but thank you to everyone who chimed in, especially @zweli !
2 Likes